ASL Fingerspelling Classification for Use in Robot Control

Abstract

This paper proposes a gesture-based control system for industrial robots. To achieve that goal, the performance of an image classifier trained on three different American Sign Language (ASL) fingerspelling image datasets is considered. Then, the three are combined into a single larger dataset, and the classifier is trained on that. The result of this process is then compared with the original three.

Keywords

• sign language
• machine vision
• convolutional neural networks
• visual communication